The development of drugs is an extremely complex and lengthy process due to stringent regulations, thorough testing, and the need for substantial investments. Financial resources and ethical considerations are the main reasons why the industry is currently seeking alternative ways to accelerate the longest initial phase of development. The use of artificial intelligence and machine learning, which have significantly advanced in the last decade, seems to be the right choice for this. Certain models allow for the prediction of the behavior of chemical compounds from drugs with targeted proteins in humans using 2D recorded data. One such model is the DEEPScreen classifier, which has demonstrated exceptionally high accuracy in its work so far. This paper fully explores the DEEPScreen classification. It covers everything from the theoretical explanation of its operation to testing on available data from the ChEMBL v23 database. The tests conducted yielded precision, recall, accuracy, F1-score, and MCC coefficient as evaluation metrics, as well as losses during training. Training was conducted with different datasets and with the same dataset but varying the number of training epochs. All results were ultimately compared, and conclusions were drawn about the performance of the developed model. The application of the DEEPScreen classifier in drug development has proven to be very successful. Based on the training results, which indicate good prediction, the use of the DEEPScreen classifier can easily accelerate the early drug development process, benefiting the entire pharmaceutical industry and its users.